1 |
On 2017-04-24 14:44, Mick wrote: |
2 |
|
3 |
> 1. Emerge the new gcc: |
4 |
> |
5 |
> emerge -uaDv gcc |
6 |
> |
7 |
> 2. List available gcc versions: |
8 |
> |
9 |
> gcc-config -l |
10 |
> [1] x86_64-pc-linux-gnu-4.9.4 * |
11 |
> [2] x86_64-pc-linux-gnu-5.4.0 |
12 |
> |
13 |
> 3. Select the new gcc version: |
14 |
> |
15 |
> gcc-config 2 |
16 |
> gcc-config -l |
17 |
> [1] x86_64-pc-linux-gnu-4.9.4 |
18 |
> [2] x86_64-pc-linux-gnu-5.4.0 * |
19 |
> |
20 |
> 4. Source the new profile: |
21 |
> |
22 |
> . /etc/profile |
23 |
> |
24 |
> 5. Rebuild all packages that have a dependency on libstdc++.so.6': |
25 |
> |
26 |
> revdep-rebuild --library 'libstdc++.so.6' -- --exclude gcc |
27 |
> |
28 |
> 6. Clean old versions of gcc you no longer need: |
29 |
> |
30 |
> emerge --depclean -a -v |
31 |
|
32 |
And now I know why I was afraid of this. It was there in the back of my |
33 |
mind, but stack height was too large. |
34 |
|
35 |
Quoting palemoon-*.eclass: |
36 |
|
37 |
gcc-5_error() { |
38 |
eerror "You are currently on a gcc 5.* compiler profile." |
39 |
eerror "Building Pale Moon with gcc >=5 results in a very unstable build." |
40 |
eerror "You can use gcc-config to change your compiler profile," |
41 |
eerror "just remember to change it back afterwards." |
42 |
eerror "You need to have the appropriate versions of gcc installed" |
43 |
eerror "for them to be shown in gcc-config." |
44 |
} |
45 |
|
46 |
And sure enough, palemoon is one of the packages revdep-rebuild wants to |
47 |
make over. |
48 |
|
49 |
So - what to do? I can't live without palemoon now. Does it make sense |
50 |
to exclude it (as in '-- --exclude gcc palemoon') and keep the old gcc |
51 |
slot around? I guess that will work if palemoon is linked with the |
52 |
-rpath option. |
53 |
|
54 |
Or do I have to mask gcc-5 until palemoon is fixed (and will it ever be?) |
55 |
|
56 |
-- |
57 |
Please *no* private Cc: on mailing lists and newsgroups |
58 |
Personal signed mail: please _encrypt_ and sign |
59 |
Don't clear-text sign: |
60 |
http://primate.net/~itz/blog/the-problem-with-gpg-signatures.html |